Approaches to Assay Processing Package Validation
Ellis Hughes
Fred Hutch Cancer Research Center - SCHARP - VISC
rstudio::conf(2020L)
Outline
- Fred Hutch and SCHARP
- Assay Pipelines
- Validation
Fred Hutch and SCHARP
Blood Stem Cell Transplantation
and Immunotherapy
Cancer Risk Factors, and Causes,
Prevention and Outcomes
Vaccine Development
and Virus-Associated Cancers
Molecular Underpinnings of Cancer
Tumor-Specific Translational Research
- Established 1992
- Worldwide-impact in the fight against cancer,
HIV/AIDS and other infectious diseases. - Clinical Data Management
- Laboratory Data Management
- Statistical Support and Analysis
Assay Processing Pipelines
image source: https://storage.needpix.com/rsynced_images/crowd-2045498_1280.jpg
Validation
Write Specifications
Code
Design Test Cases
Evaluate Test Cases
Document Results for Final Sign Off
Specifications
Do:
- Set the goals of the project
- Define current state
- Specific Features
- System agnostic
- Define success
- Set a time line
Dont:
- Leave ambiguity final product
- Define exact process
- Leave Success metrics up to interpretation
Documentation
- History of Specifications
- Specifications
- Measureables
- success
#' @title Specifications For RStudio Conf 2020 Success
#' @section Last updated by:
#' Ellis Hughes
#' @section Last updated date:
#' 2020-1-27
#'
+ _Specifications_
+ 1.1 Presentation must explain validation procedure.
+ 1.2 Be entertaining by causing 3 laughter sessions.
+ 1.3 Inform and document each step necessary for success.
+ 1.4 (Optional) Fame and Glamour and start branded accessories chain.
Code
Step One
- Define function requirements
- Design structure
- Measurable goals
- Distribute workload
- Modularity
- Update as necessary
- Break out of ‘validated’ processing pipelines as necessary
Supporting Cast
-
Roxygen Documentation
WHAT, WHERE, WHY, HOW
WHO, WHEN
Unit Tests
Advantages of SCM
Branching
-
Detailed code Review
- Style guides
Documentation
#' @title Function Name
#' @description
#' Description of great function
#' @section Last updated by:
#' Ellis Hughes
#' @section Last updated date:
#' 2020-1-27
#' @param arg1 argument 1
#' @param arg2 argument 2
#' @example
#' output <- sample_function(1,2)
#'
#' @export
sample_function <- function(arg1, arg2){
temp_object <- arg1 * arg2
return(temp_object)
}
Test Cases
What are they
Not Unit or Regression tests
-
Draw connections between specifications and code
- Prove Code satisfied Specifications
The Approach
Detailed without writing code
How do all the pieces actually work together
Test Case Coding
A case for Test Coding
Resolves interpretation errors in documentation and test cases
-
3rd party to code
- Does not have preconceived idea of function
Identify improvements
Formatting
Test that format
Combine with Roxygen comments
-
Breaking out
- each family of tests into a file
- each sub-test as a testthat case
Final Document
Record All elements in a single location
-
Following good programming practices led to
- Scraping functions for author/edits
- separating out specs into files allows for modular extension
- Roxygen to document editing and edit dates
- using testthat like structure led to ability to using testthat reports to document success/failures
- Automation is mo betta
Final Document
Many thanks to all involved
Fred Hutch, Managers, and Team
